Embodiment, Semantics and Social Action: The Case of Object-Transfer in L2 Classroom Interaction

Using conversation analysis and usage-based linguistics, I focus on a beginning L2 user in an ESL classroom and trace his use of a “family of expressions” which, from the perspective of linguistic theory, are instantiations of either the ditransitive dative construction (e.g., “he told me the story”) or a prepositional dative construction (e.g., “he told the story to me”). The semantics of both constructions denotes transfer of an object, physically or metaphorically, from one agent to another. Therefore, I investigate them as one type of object-transfer construction. The instances of the construction are found predominantly in instruction sequences, and I show how the L2 user co-employs talk and recycled embodied work that elaborates the deictic references of the talk and the relation of agent-object-recipient roles among them. Through my analyses, I will showcase the embodied nature of linguistic categorization (Langacker, 1987) but take the argument further and suggest that the semiotic resource known as “language” is a residual of embodied social sense-making practices (aus der Wieschen and Eskildsen, 2019). The study draws on the MAELC database at Portland State University, a longitudinal audio-visual corpus of American English L2 classroom interaction.

Qualifications of Instruction Sequence Failures, Faults and Defects: Dormant, Effective, Detected, Temporary, and Permanent

Starting out from the survey of instruction sequence faults from [6] program faults are classified according to the conventional criteria of being dormant, effective, detected, temporary, and permanent. Being retrospectively approved is introduced as an additional qualification. For this theoretical investigation the context is simplified by contemplating instruction sequences as a theoretical model for programs, and by assuming that instruction sequences are supposed to compute total transformations on finite bit sequences of a fixed length only. The main conclusion which can be drawn from this work concerns the notion of dormancy. First of all it is noticed that the unconventional notion of a dormant failure is both plausible and amenable to a straightforward and convincing definition. The conventional notion of a dormant fault, however, is much harder to grasp and the definition of a dormant fault which is provided in the paper may be disputed. The notion of a dormant fault seems to admit no convincing intuition. All faults are defects but not the other way around. The idea of a fault exclusively depends on an instruction sequence and a specification of which it is considered to be a candidate implementation. In the presence of a design, however, in addition to faults, the notion of a deviation from design (DFD) defect arises, which constitutes a class of defects many of which are not faults. For DFD defects the notion of dormancy admits a straightforward and convincing definition.

Data Efficient Algorithms and Interpretability Requirements for Personalized Assessment of Taskable AI Systems

The vast diversity of internal designs of black-box AI systems and their nuanced zones of safe functionality make it difficult for a layperson to use them without unintended side effects. The focus of my dissertation is to develop algorithms and requirements of interpretability that would enable a user to assess and understand the limits of an AI system's safe operability. We develop an assessment module that lets an AI system execute high-level instruction sequences in simulators and answer the user queries about its execution of sequences of actions. Our results show that such a primitive query-response capability is sufficient to efficiently derive a user-interpretable model of the system in stationary, fully observable, and deterministic settings.

Instruction Sequence Faults with Formal Change Justification

The notion of an instruction sequence fault is considered as a theoretical concept, for which the justification of the qualification of a fragment as faulty is mathematical instead of pragmatic, the latter approach being much more common. Starting from so-called Laski faults a range of patterns of faults and changes thereof for instruction sequences is developed.

Exploring the effects of demonstration and enactment in facilitating recall of instructions in working memory

Across the lifespan the ability to follow instructions is essential for the successful completion of a multitude of daily activities. This ability will often rely on the storage and processing of information in working memory, and previous research in this domain has found that self-enactment at encoding or observing other-enactment at encoding (demonstration) improves performance at recall. However, no working memory research has directly compared these manipulations. Experiment 1 explored the effects of both self-enactment and demonstration on young adults’ (N=48) recall of action-object instruction sequences (e.g. ‘spin the circle, tap the square’). Both manipulations improved recall, with demonstration providing relatively larger boosts to performance across conditions. More detailed analyses suggested that this improvement was driven by improving the representations of actions, rather than objects, in these action-object sequences. Experiment 2 (N=24) explored this further, removing the objects from the physical environment and comparing partial demonstration (i.e. action-only or object-only) with no or full demonstration. The results showed that partial demonstration only benefitted features that were demonstrated, while full demonstration improved memory for actions, objects and their pairings. Overall these experiments indicate how self-enactment, and particularly demonstration, can benefit verbal recall of instruction sequences through the engagement of visuo-motor processes that provide additional forms of coding to support working memory performance.